2023 Volume 64 Issue 7 Pages 1464-1473
Recovery of cold worked metals is associated with a loss of strength due to the reduction of the defect density. However, already in the 1960s analytical models and few experiments suggested that this may not generally be the case and even a hardening rather than a softening might occur. With the availability of severe plastic deformation and deposition techniques this anneal hardening phenomenon has been observed frequently. In this overview we summarize early findings on this topic before focusing on general observations and potential origins, with a special focus to show the similarities for nanostructures across different grain size scales (i.e., structures prepared by deposition and severe plastic deformation techniques, respectively). Comparison of different results indicate a grain size dependent hardening increment that could be additionally affected by segregation or the processing variables. Considering the agreement of the peak hardening temperatures with that for dislocation annihilation at grain boundaries, anneal hardening can be rationalized by the loss of intragranular defects and grain boundary relaxation. Grain boundary diffusivity hence plays a crucial role and particular solutes could amplify the hardening process even further. As anneal hardening already occurs for grain sizes at the micron scale, its effect on properties is even of technological relevance. Beneficial effects on the fatigue strength are evident, but the strain softening of anneal hardened specimens drastically shorten ductility. Nevertheless, some strategies to overcome this adverse effect seem promising to create metallic structures with exceptional combinations of strength and ductility.
Although during cold working of metals only a small portion (<10%) of the spent energy is stored within the microstructure, the material has a tendency to recover upon subsequent annealing, i.e., to return to the defect-scarce state.1–3) Accordingly, properties and microstructures are subject to change and thus recovery and related annealing processes such as subgrain growth or recrystallization have been studied extensively for almost a century.3–6) Understanding and predicting these processes is crucial to design microstructures with desired properties or crystallographic textures.7,8) Upon recovery point defects and redundant dislocations annihilate and dislocation tangles or dislocation cells reorder into lower energy configurations, possibly accompanied by subgrain growth. Recovery may also occur at grain boundaries, which could be transferred into a higher energy state during deformation, i.e., contain excess defects.9,10) Hence, the recovered state with its lower defect density is expected to have a reduced strength compared to the as-cold worked condition. However, already in 1968 Ashby proposed on a purely analytical basis that the stresses required to multiply dislocations in well-annealed crystals (i.e., corresponding to the flow stress after a recovery heat-treatment) may not generally decrease compared to the initial state (i.e., the cold-worked condition).11) This hypothesis was based on the fact that a dislocation segment can only act as a source if it lies within a slip plane and also contains its Burgers vector. The probability to meet this condition is likely for medium or high dislocation densities, but then the average segment length is small, leading to large multiplication stresses. In contrast, if the density falls below a certain value, despite a large segment length, the probability to act as a source becomes also low, what results as well in larger stresses to multiply dislocations. This prediction has meanwhile been reproduced by discrete dislocation dynamics simulations.12) The generally expected and observed softening upon recovery could thus change into a hardening upon recovery for specific material states. Such condition could be achieved if either the dislocation density is already very low after synthesis or could be effectively reduced due to heat treatments. This is indeed found experimentally for defect scarce, as-cast microwires,13) testing of small sample volumes where the probability to probe existing dislocation sources is low (e.g., prepared by ion-ablation14)), but also for polycrystals consisting of very fine grains and thus large fractions of extremely effective dislocation sinks. Although dislocation storage is already less likely in submicron or nanoscale grains, annealing may further reduce this probability and the lack of existing mobile dislocation segments results in an enhancement of the yield strength, i.e., a hardening by recovery.
Hardening during recovery annealing treatments has indeed been observed for various submicron sized or nanocrystalline (NC) materials. The frequent availability of deposition techniques and top-down approaches such as severe plastic deformation (SPD), made this phenomenon meanwhile to a rather common observation (compare for example Refs. 15)–24)). However, as these observations were made on specimens spanning over almost two orders of magnitude in grain sizes, involving different purity levels and/or solute contents, the presence or absence of second phases, the magnitude of the anneal hardening effect was found to differ largely. While in some cases the effect is hardly measurable or even the classic softening is observed,22,25) for other nanostructures like Fe–Ni26) or Ni–Mo21) recovery can almost double the strength, even compensating the measured inverse Hall-Petch effect for the as-prepared specimens.21) The underlying processes enabling this extra strengthening are hence still a matter of debate. Undoubtedly, thorough understanding could allow to fully exploit the use of this extra strengthening mechanism, eventually enabling the design of metals with close to theoretical strength. With this short overview article, we aim to sum up early reports on this topic, to describe some common observations of the anneal hardening phenomenon, but also to focus on recent results which might allow to improve our understanding regarding the fundamental processes triggering the strengthening potential of nanomaterials during recovery treatments. Moreover, we will present results emphasizing that apart from NC materials, the effect could even be important and effective to alter properties of technically relevant microstructures, i.e., around the micronscale or for technically feasible degrees of cold working. It should be emphasized that this article deliberately focuses on the anneal hardening effect in single-phase metallic materials where precipitation or phase separation remain absent during the recovery anneal.
Apart from the theoretical prediction of Ashby that a reduction of the dislocation density does not necessarily lower the strength of metals, first reports of a recovery hardening instead of the expected softening date back to the late 1950s and 1960s.27–29) Hardness of heavily cold rolled brass and Cu–Al alloys was found to increase prior to recrystallization. This unexpected increase was rationalized by segregation of solute to the stacking faults of extended dislocations, presumably accompanied with the formation of chemical short range order.27–29) However, as will be discussed in the subsequent sections, the low stacking fault energy of these alloys could also play a distinct role for the measured hardness increase during recovery. A low stacking fault energy favors mechanical twinning and hence extremely efficient microstructural refinement already for comparably small applied strains. Although for the mentioned cold rolled brass and bronze samples recovery hardening was attributed to solute segregation and/or the formation of chemical short range order,27–29) a significantly reduced microstructural length scale seems to be a further prerequisite to observe the anneal hardening phenomenon - with the invention of NC metals by deposition30,31) or SPD techniques32–34) almost four decades ago the anneal hardening phenomenon has become a common observation.
Together with suitable annealing treatments these techniques allowed to adjust even for pure metals grain sizes at the micron scale and to investigate their properties. As an example severe extrusion and slight recrystallization treatments gave access to pure aluminum samples with very small grain sizes of 2–14 µm.35) Studying the properties of these fine-grained aluminum specimens using uniaxial tensile tests, Polish scientists made a striking observation. For the finest grain sizes yield point phenomena, i.e., Lüders bands, were recorded, with the Lüders strain increasing for smaller grain sizes, see Fig. 1. In contrast, specimens with 14 µm grain size showed continuous yielding, as expected for pure aluminum, Fig. 1.
Load-elongation curves obtained at 293 K on pure aluminum (99.99%) samples with different grain sizes. With a reduction in grain size Lüders deformation becomes increasingly dominant. Reprinted with permission from Ref. 35).
Another intriguing finding is that the Lüders deformation was only observed for the fine-grained high-purity aluminum (99.99%), but not for a batch of specimens with somewhat lower purity (99.7%) but similar grain sizes.35) The lack of mobile dislocations and dislocation sources in the annealed specimens (i.e., the grain sizes were adjusted by different recrystallization treatments) were suggested to cause the Lüders deformation. This indicates that the Lüders phenomenon is not necessarily related to solute or impurity atoms, but can be a consequence of sufficiently fine and defect scarce grains. This has meanwhile been proven for a variety of materials36–39) – independent of the studied system, Lüders deformation was generally observed for specimens consisting of submicron or few micron sized grains prepared by adequate heat treatment of severely deformed specimens. An increase of the hardness or yield stress was not observed in these studies, what may not be surprising as the grain size adjusted by heat treatments was considerably larger than in the as cold worked state. However, it seems that in some cases the onset of microplasticity occurs at higher stress levels. This indicates that already for micrometer sized grains properties could be largely dictated by the amount of intra- and intergranular defects and not solely by the grain size. Further studies conducted by the Polish group clearly point into this direction. Depending on the amount of grain boundary defects, the Hall-Petch constant was found to differ by a factor of five.40) For even finer grain sizes, or when as-cold worked and annealed conditions of similar grain size could be compared, the described phenomenon might become intensified. This is indeed the case and Huang and co-workers noticed a 9% increase of the yield stress of a 200 nm single phase, pure aluminum sample processed by accumulative roll bonding (ARB).15) The frequent availability of submicron sized or even NC metals by deposition or SPD techniques allowed indeed for frequent observations of a hardness or strength increase upon recovery, rather than the observation of Lüders bands. It should be noted, that already long-term storage of nanostructures at ambient temperature could induce a slight hardness increase, with the magnitude depending on the purity and melting point of the sample.41) Different origins for the anneal hardening and the appearance of Lüders bands have been proposed, although they might have a common origin. General observations, potential origins and findings of recent studies focusing on the anneal hardening phenomenon shall be discussed in the following chapters.
2.2 Observations on ultra-fine grained or nanocrystalline metalsAs noted above, the grain size seems to play a distinct role for the anneal hardening phenomenon. Although SPD can refine even the grain size of pure metals to a few hundred nanometers, a hardness increase due to recovery treatments may not be generally observed. Two examples, pure nickel (99.99%) and an Al1Mg alloy deformed by high-pressure torsion (HPT) at ambient temperature and subsequently subjected to isochronal annealing for 30 minutes are shown in Fig. 2. The grain sizes after HPT were 240 nm for the pure nickel and about 200 nm for the Al1Mg samples.22,42) Prior to a pronounced drop in hardness, related to the onset of significant grain growth, hardness did not change significantly or even reduced slightly in case of pure nickel. In contrast, for nickel samples with slightly lower purity level (99.6%, ∼100 nm grain size) a small hardness increase could be measured,25) which was even more pronounced for a Ni4.5Al alloy with about 120 nm grain size,22) Fig. 2. As increasing solute or impurity content reduce the achievable grain size upon SPD, this suggests a dependence of the extent of anneal hardening, ΔH, on the grain size, but also that solute/impurity atoms could play an additional role. Indeed, several studies indicate a clear dependence between ΔH and the grain size – the smaller the as-prepared grain size, the larger the hardening increment, see Fig. 2(b). Especially for nanomaterials with extremely fine grain sizes (<20 nm) prepared by deposition techniques, where often an inverse Hall-Petch behavior is reported (i.e., a reduction in strength towards finer grain sizes), the magnitude of the hardening can even overcompensate the softening from the inverse Hall-Petch effect measured in the as-deposited state.21) Such continuous strengthening down to grain sizes of only a few nanometers may however not be observed for any annealed nanocrystalline state, compare Refs. 24), 43). Moreover, the data presented in Fig. 2 indicates that a critical grain size may exist, which separates the classic recovery softening from recovery hardening. A sufficiently fine grain size and solute or impurity atoms may thus be necessary to observe anneal hardening. However, this interpretation could be too simplified, as discussed in the following.
(a) Room temperature hardness of various nanostructured metals after isochronal annealing. Despite grain sizes smaller than 250 nm, a hardness increase after recovery is not observed in general. Data is taken from Refs. 22), 25), 42). (b) Extent of anneal hardening, ΔH, as a function of the initial grain size obtained on different systems. The data suggests a pronounced grain size dependence of ΔH. Data is taken from Refs. 18), 21), 44).
Separating grain size from solute effects remains extremely challenging, as for a particular purity level or chosen composition of the alloy the grain size can only be varied by adapting the SPD processing temperature or the deposition parameters (e.g., change of the substrate temperature etc.). Adjusting these process variables could, however, already affect the defect content and hence the extent of recovery during subsequent ageing. On the other hand, for particular synthesis parameters grain size could only be varied by changing the chemical composition. This changes the amount of solute involved and could also affect the defect densities and recovery behavior. We shall discuss this pending issue in detail in the section on the underlying processes governing anneal hardening.
In case of recovery hardening, hardness of the annealed specimens increases up to a peak annealing temperature, Tpeak, above which distinct grain growth occurs. Independent of the annealing temperature, a rapid hardening can be observed and prolonged annealing times result only in a slight additional hardness change, Fig. 3(a). Deviations from such hardness plateau and further strengthening with prolonged annealing times might be a hint for precipitation or phase decomposition, as for instance observed for NC HEAs.45) It should be noted that already for annealing at the peak temperature, local microstructural changes can occur, while the average grain size may only shift slightly. Above the peak temperature grain growth becomes significant and hardness drops accordingly. Still, hardness could remain at levels being higher or comparable to the finer-grained as-prepared condition, see Fig. 3(b). This indicates that despite a potentially pronounced bimodality of the microstructure a sufficiently large fraction of fine crystallites prevailed, and/or that even grains of larger size require a higher stress level to be plastically deformed. This is in line with the mentioned investigation from the 1980s, emphasizing that the Hall-Petch constant is extremely sensitive to the processing history and could vary largely for the same grain size.
Isothermal annealing treatments indicate a rapid hardening before a constant plateau hardness is reached. Deviations from the constant plateau may indicate, as for the NC HEA, precipitation or decomposition processes. Data is taken from Refs. 18), 20), 44), 45). (b) Typical anneal hardening measured on a NC Pt10Ru alloy deformed by HPT at 77 K. Grain size distributions of circled conditions are presented in the diagram to the right. Despite significant grain growth occurs above the peak temperature, hardness can still reach similar values as for the as-deformed state. Figure is adapted from Ref. 44).
While the critical temperature where grain growth becomes significant depends on the amount and type of solute, the grain size and synthesis method of the nanostructure seem to play only a minor role, see Fig. 4. In Fig. 4 hardness values after isochronal annealing treatments of nanostructured nickel22,46,47) and Ni–Mo alloys21,48) prepared by HPT and electrodeposition, are compared. Note that for a particular chemistry electrodeposition generally results in much finer grain sizes than accessible by SPD techniques. As an example, depending on the purity level, HPT deformed pure nickel has a grain size of 100–300 nm,22,25,49) while a significantly finer grain size of 20–30 nm is accessible by electrodeposition.47,50) Data obtained on HPT deformed Ni powder,46,51) having a significantly finer grain size than the bulk HPT deformed nickel samples, are displayed for comparison.
Hardness of HPT deformed (bulk nickel and consolidated nickel powder) and electrodeposited (ED) nanostructured nickel and Ni–Mo alloys as a function of the annealing temperature. The temperature to induce a drop in hardness because of the onset of significant grain growth is rather independent of the initial grain size and processing technique. Data is taken from Refs. 21), 22), 42), 46)–48). Trendlines for the Ni–Mo alloys, the ED Ni and the powder based NC Ni are displayed just as a guide for the eye.
Although the grain sizes differed by up to an order of magnitude, the peak temperatures for grain coarsening (i.e., the temperature where hardness starts to drop), remained essentially unchanged. However, the extent of anneal hardening, ΔH, was much larger in case of the electrodeposited samples which consist of much finer grains. This is in line with the observed grain size dependence of ΔH. Extrapolation of the grain size dependence of the ΔH values for a particular material further suggests that a certain grain size exist where a crossover from recovery hardening to the classic softening occurs. Studies on nanostructured Pt–Ru44) and Ni–Mo alloys,21) suggest that zero hardening could be expected for grain sizes of a few hundred nanometers. In contrast, one of the first reports on anneal hardening investigated aluminum processed by ARB, with grain sizes exactly in this range (∼200 nm).15) The expected transition from recovery hardening to the classic softening may thus depend on the material, but also the testing method could explain the afore mentioned discrepancy, as follows. The ARB processed aluminum samples were tested in uniaxial tension while for the HPT deformed Pt–Ru and electrodeposited Ni–Mo samples their limited sample volumes only allowed for hardness measurements. Using self-similar indenter tips such as a Vickers pyramid or a Berkovich tip already rather large equivalent strains (about 8% for a Vickers or Berkovich tip) are imposed.52) Such strain levels are typically much larger than the uniform elongation of a nanocrystalline material in a tensile test. Considering that recovery of the as-deformed or deposited states causes a strength increase, deformation after the heat treatment will gradually soften the material towards the initial strength level, i.e., the recovered conditions will strain soften.15,44) For a particular sample state, the increase of the yield or ultimate tensile strength (for nanomaterials already reached for strains of a few percent) upon annealing could thus be higher than expected from hardness data. This is an important fact which should be taken into account when data of different experiments are analyzed. A comparison of the anneal hardening increment, ΔH, measured using two different indenter shapes, a Berkovich and a cube corner tip applying 8% and 20% equivalent strain, respectively, shows a clear reduction of ΔH with increasing applied strain, compare Fig. 5. Hence, zero hardening increment measured on recovered samples using self-similar indenters does not necessarily imply that the anneal hardening phenomenon is not active. The transition from recovery hardening to softening will thus not occur at grain sizes of a few 100 nm, but already at even larger grain sizes. The appearance of Lüders deformation for recovered or recrystallized materials having grain sizes slightly below or above the micron scale support this view.
Hardness of NC Pt10Ru samples (synthesized by HPT at 77 K) after different annealing treatments measured with two different indenter tips (i.e., at different equivalent strains), a Berkovich and cube corner tip, respectively. Pronounced strain softening of the annealed conditions is evident. Reprinted with permission from Ref. 44).
Although not in the focus of the present work, similar hardening mechanisms could also occur for (nano)composite materials, see for instance results on severely cold-drawn pearlitic steels or HPT deformed Cu–Cr composites,53,54) where for slight annealing hardness and strength tended to increase. However, the mechanisms enabling the extra strengthening could be more complex and could differ from single phase materials (which will be discussed in the following section), involving in addition ordering or precipitation processes resulting from mechanical mixing that occurred during severe cold deformation.
As briefly mentioned in the last paragraph, the extra strengthening of nanometals upon subsequent annealing could have multiple origins, but only those playing a role for single-phase metals will be discussed in the remainder. In the absence of second phase formation and distinct grain growth, only the densities of point and line defects will be affected during the heat treatment. Thus, a change of hardness or strength compared to the as-prepared condition needs to be attributed to these changes. This may not only involve a reduction of the excess defect densities within the grains and at grain boundaries (i.e., grain boundary relaxation) but also additional changes such as solute segregation to grain boundaries. Detailed studies on ARB processed aluminum revealed a clear reduction of the dislocation density by a factor of two during a recovery anneal.15) The pre-existing dislocations that facilitate early yielding in the as-ARB state need to be nucleated after the recovery treatment.15,16) As for submicron and especially NC metals the probability for intragranular dislocation sources diminishes, initiating plasticity requires the activation of grain boundary sources. But also the activation stress for these boundary sources increases due to the annealing treatment as excess defects within the boundaries gradually disappear, as shown in atomistic simulations.55) The annihilation of mobile (‘ready to go’) dislocation debris and the relaxation of grain boundaries, thereby hardening the dislocation sources, was thus suggested to be responsible for the recovery hardening.15,16,18,20,22)
Apart from dislocation annihilation at grain boundaries and their subsequent relaxation, also the formation of vacancy agglomerates has been suggested to cause the anneal hardening phenomenon in some studies.56,57) Large excess vacancy concentrations (cvac ∼ 10−5–10−4) have been measured indirectly in severely deformed metals.58) Their coalescence during annealing could lead to the formation of dislocation loops which increase the flow stress by an Orowan mechanism.59–61) It should be noted, that the increase in hardness or flow stress due to vacancy agglomerates diminishes with applied strain, i.e., reduced crystallite size.56,57) This might be related to the fact that grain boundaries can act as sinks for (point) defects and vacancy agglomerates can be largely avoided for smaller grain sizes. This makes nanostructured materials extremely irradiation tolerant.62,63) The potential contribution of vacancy clusters to the anneal hardening should thus diminish with decreasing grain size - the opposite to what is observed. For this reason, we do not consider the contribution of vacancy agglomerates to the anneal hardening phenomenon as the most relevant. However, for coarse grained specimens, especially those containing huge excess concentrations of vacancies (e.g., quenched-in vacancies), an effect is clearly measurable. This is especially the case for hcp metals where loop formation predominately occurs at the basal planes.59)
That the annihilation of dislocations at grain boundaries and their relaxation are the more relevant processes to consider is also evident from a recent study. There, for the same materials, the peak temperatures for anneal hardening of the severely deformed condition were compared to those observed for dislocation annihilation in random high-angle grain boundaries of a coarse grained state, compare Refs. 22), 64) and Table 1. The nanostructures were produced by HPT and consisted predominately of random high-angle grain boundaries, making the two conditions comparable. By subsequently applying isochronal annealing treatments the peak temperatures separating recovery (either hardening or softening was observed) from subsequent grain growth were recorded, Table 1. Overall, the agreement between the two temperatures is striking, but reasonable at a closer look. The annihilation or accommodation of lattice dislocations at grain boundaries requires their dissociation or spreading within the boundaries,10,65) i.e., the possibility of slight movements or rearrangements, the same what would be necessary to induce grain growth. These processes can effectively reduce (relax) the grain boundary energy and have also been used to determine grain boundary diffusion coefficients.66) Models describing the relaxation time, τ, (eq. (1), Refs. 65), 67)) reflect this dependence on the grain boundary diffusivity, DGB.
| \begin{equation} \tau = \frac{k_{B}Ts^{3}}{G\varOmega \delta D_{\textit{GB}}} \end{equation} | (1) |
The discussion above is further reflected in the close match of peak temperatures (Fig. 4), independent of the synthesis method or grain size. Even for nickel, HPT deformed at 77 K, where a huge density of grain boundary excess defects can be assumed, the peak temperature only changed by about 35 K, Fig. 4. However, as the boundary diffusivity plays an important role also solute excess at grain boundaries is expected to affect the extra strengthening by recovery annealing. This was indeed suggested from experiments17,19,69) and atomistic simulations.70) Nevertheless, an active role of the segregated solute in terms of a pinning-unpinning of dislocations, enhancing the overall stress level for their propagation, is not expected. This should result in negative strain rate sensitivities of the flow stress, which are experimentally not observed, Fig. 6.
(a) Hardness of NC Pt10Ru synthesized by HPT at 77 K obtained with a Berkovich tip at a constant strain rate of 0.05 s−1 as a function of the testing temperature. Measurements obtained during cooling clearly reflect the anneal hardening. (b) Results of strain rate jump tests at different indentation temperatures. For all testing temperatures a positive rate sensitivity is evident. Reprinted with permission from Ref. 44).
The enormous potential to strengthen nanomaterials by proper interfacial design is clearly evident from a comparison of strength of different aluminum alloys,71–74) Fig. 7. From a Hall-Petch type plot, depending on the type of solute, for a given grain size quite different strength levels can be deduced, compare Fig. 7. As an example, when using rare earth alloying elements (e.g., La and Ce), the strength levels off or even decreases from about 800 MPa when reducing the grain size below 100 nm. This is in contrast to commercial alloys (2000, 5000 and 7000 series), were much higher strength levels can be obtained for the same grain size. Understanding the role of segregation effects could thus help to specifically tailor and exploit the anneal hardening potential of nanostructures. Nevertheless, attempts to decouple intrinsic grain size from solute or doping effects remain challenging, as the grain size can either be adjusted by a change of the processing parameters or a change of the chemical composition. A change of the processing parameters such as the substrate temperature or the SPD deformation temperature could lead however, already to a significant change of the defect content, thus different strength values.
Hall-Petch type plot of different nanostructured aluminum alloys. For a given grain size, different alloying concepts result in different strength levels. Rare earth alloying elements result in rather low strength levels. This suggests an important role of solute on the strength of nanomaterials. The figure is adapted from Ref. 71) with data from Refs. 72)–74).
Especially for very fine grained nanometals (i.e., grain sizes less than ∼20 nm) another origin of the anneal hardening is often suggested. Heat treatments provoking solute segregation and grain boundary relaxation are proposed to prevent a shift of the deformation mechanisms from dislocation to grain boundary mediated processes (e.g., such as grain rotation, sliding or migration), thought to be responsible for the frequently observed inverse Hall-Petch behavior.21) If a change of the deformation mechanism takes place is still a matter of debate, but several aspects suggest that this is not the case. First, even for sub-10 nm grain sizes prepared by cryogenic sliding,75) dislocations were found within the grains. Moreover, the normalized probability distribution of the boundary spacings was identical to that of cold-worked coarse grained metals, emphasizing that these nanograined structures evolved by dislocation plasticity. Second, grain boundary based processes are already active during plastic deformation at much larger grain sizes (e.g., 200 nm),76–80) where formed deformation textures also provide a clear finger print for dislocation based plasticity. However, deformation induced boundary migration over a few nanometers will not dramatically affect properties for grain sizes of a few 100 nm but certainly for the finest grain sizes. Lastly, the peak temperatures between different grain sizes do not change significantly and perfectly match with those measured for dislocation annihilation, Fig. 4. Therefore, one could argue that the origin of anneal hardening will not change significantly over the discussed length scales. However, the probability of dislocation storage within the grains will diminish and dislocations will only be emitted from grain boundaries, propagate through the grain, before becoming absorbed at the opposite boundary. These dislocation processes and the required stresses to operate them could certainly be affected by heat treatments. For smaller grain sizes recovery could thus lead to grains and grain boundaries with much lower defect contents, causing higher stresses to emit but also absorb dislocations, reflected in more pronounced property changes.
While recovery can have exceptional effects on strength, a major consequence of anneal hardening is the occurrence of strain softening upon subsequent deformation.15,44) Strain softening favors rapid strain localization and early failure in a tensile test. Compared to the as-prepared state ductility is in many cases massively reduced, resulting often in a macroscopically brittle fracture behavior along a single shear or deformation band. Despite this brittle appearance at the macroscale the fracture is still of a microductile type, consisting of a shallow dimple structure.81) Balancing the softening for small grain sizes, where the anneal hardening becomes significant, remains challenging as work hardening from dislocation storage is widely absent. Only for grain sizes >500 nm such work hardening can balance the strain softening, resulting in the mentioned Lüders deformation.82) Nevertheless, several attempts to maintain ductility even for the finest grain sizes, while still benefitting from the exceptional anneal hardening at that scale, have been made. One of these approaches targets grain boundary phase transformations induced by heavily segregating elements.83,84) In the extreme case this structural transformation may even result in the formation of thin amorphous layers which could drastically alter the dislocation emission and absorption behavior of the grain boundaries, claimed to preserve ductility.85) As these transformations require a rather complex thermomechanical processing route and reproducing studies failed to form these amorphous films,86) more facile approaches need to be targeted.
Promising findings in this direction were recently published. Nanostructured aluminum alloys and NC austenitic steel were subjected to a recovery heat treatment to provoke anneal hardening. Subsequently, both HPT specimens were subjected to a slight additional HPT increment at ambient temperature.87,88) As expected, for both materials strength was slightly reduced compared to the as-recovered condition, but still remained higher than the as-HPT deformed state. Despite the higher strength, ductility was as high or even enlarged compared to the as-HPT deformed condition,87,88) Fig. 8(a). Why even higher strains to failure compared to the as-HPT processed condition could be measured requires further in-depth research, what would enable to tune this promising approach towards unprecedented combinations of exceptional strength and ductility.
(a) Comparison of the tensile stress-strain curves obtained on HPT deformed NC austenitic steel samples, anneal hardened batches and samples which were additionally deformed after the relaxation annealing treatment. Slight deformation after the heat treatment allows for exceptional combinations of strength and ductility. (b) Effect of anneal hardening on the fatigue limit of NC austenitic steel, enabling a further 60% increase compared to the as-HPT deformed condition. Figure is reprinted with permission from Refs. 81), 87).
Apart from these attempts to overcome the strength-ductility dilemma of anneal hardened conditions, even this recovered state could be of interest if exceptional fatigue strength is required. As fatigue induced grain coarsening of nanostructures was identified to be even in the high cycle fatigue regime the initial step for fatigue crack initiation,89,90) and the fatigue induced growth process depends on the cyclic (micro)plastic strain amplitude,91,92) the fatigue limit of a recovered NC specimen should be significantly improved. Recovery removes mobile dislocations, relaxes grain boundaries and could induce solute segregation, what shifts the onset stresses for microplasticity to much higher stress levels. Grain coarsening and subsequent fatigue crack initiation within these coarsened regions could thus be effectively retarded. This is indeed the case and for an anneal hardened NC 316L steel the fatigue limit could be improved by additional 60% compared to the as-HPT deformed state, reaching almost 1 GPa for a stress ratio of R = −1, compare Fig. 8(b). Subduing fatigue induced grain growth and a reduced crack growth rate of the recovered condition seem to be responsible for this exceptional fatigue strength.81,93) Results obtained on nanostructured nickel with ∼170 nm grain size, demonstrating a 10% increase of the fatigue strength for the recovered condition94) suggest a general applicability of this strategy to develop extremely fatigue resistant metallic structures. It should be noted that the positive effect of anneal hardening on the fatigue strength is more pronounced than that on quasi-static strength. As an example, for the recovered NC austenitic steel strength was only increased by 20%, while the fatigue limit could be enhanced by another 60%, see Fig. 8(b). This indicates that already a slight anneal hardening potential could have significant effects on the fatigue strength.
As slight anneal hardening effects can already be expected for grain sizes just below a micrometer, the beneficial effects of recovery on the fatigue limit could even be exploited using technologically relevant processing methods. First, with established additive manufacturing techniques the spacing between boundaries can be rather small and even cell sizes of a few 100 nm are readily accessible.95,96) Second, for fcc metals with a low stacking fault energy such as austenitic steels or brass, mechanical twinning allows to obtain nanoscale microstructures even with classic cold forming processes (e.g., cold rolling or extrusion) and moderate equivalent strains. That the first reports of recovery hardening were observed on aluminum bronzes and brass is hence not surprising.27–29) Considering that for austenitic steels anneal hardening can already be observed for cold rolling to logarithmic strains being larger than 0.7,81) already for industrially relevant degrees of deformation improvements of the fatigue life time by annealing can be expected.
Recovery of cold worked single phase metallic structures with grain sizes at or well-below the micron scale can lead to yield point phenomena and/or further strengthening, in clear contrast to the softening effect for coarser grained structures. Recovery hardening can also be observed for nanostructures processed by other far-equilibrium processing methods such as deposition techniques. Surveying literature data indicates that independent of the processing method or the grain size regime the extent of the hardness increase is grain size dependent. However, as different grain sizes can only be adjusted by adapting the solute content or the processing parameters, these extrinsic factors may play an additional role. A comparison of temperatures required for annihilation of dislocations at grain boundaries with those leading to the maximum hardness increase upon recovery shows excellent agreement for a variety of alloys. This indicates that the annihilation of pre-existing dislocations and the subsequent relaxation of boundaries causes the anneal hardening. As these processes strongly depend on the boundary diffusivity, the boundary character and solute excess could provide a lever to even further strengthen nanomaterials. The potential can be considered huge, as recovery can double the strength of nanostructures. An important, but widely ignored fact to consider is the testing method. Hardness measurements are most often used to determine mechanical property changes. Thus, rather large equivalent strains are applied. Considering that anneal hardened structures strain soften, the actual yield strength would be even higher than expected from hardness data. The involved large number of dislocations during indentation questions if the hardness increase that is measured can still be related to the hardening of dislocation sources. However, while the anneal hardening can result in impressive additional strengthening and extraordinary fatigue limits, the strain softening upon subsequent deformation drastically shortens ductility. Recent results suggest that this shortcoming could be overcome by slight plastic deformation after the recovery anneal. This could lead to even higher ductility than in the as-prepared nanocrystalline state. Research uncovering the reasons for this unexpected behavior likely result in strategies to unite ultra-high strength with adequate ductility.
OR acknowledges funding from the Austrian Academy of Sciences via Innovation Fund project IF 2019–37.
